AI Paper Atlas

Research map / Federated learning and privacy / Privacy attacks and secure learning

Good Gradients Poison Your Model: Evading Defenses in Federated Learning via Boundary-adaptive Perturbation

Xiaojie Zhao, Jinqiao Shi, Yi Li, Junmin Huang, Chongru Fan

AAAI 2026 · Federated learning and privacy › Privacy attacks and secure learning · approach: Defending federated learning against poisoning · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use BAPerturb instead of FedPerturbmodel poisoning attack · MNIST (LeNet), defense FedAvgattack success rate (1 - test accuracy): 0.3132 -> 0.7235Improved
use BAPerturb instead of FedPerturbmodel poisoning attack · MNIST (LeNet), defense Multi-Krumattack success rate (1 - test accuracy): 0.0126 -> 0.1354Improved
use BAPerturb instead of FedPerturbmodel poisoning attack · Fashion-MNIST (CNN), defense Multi-Krumattack success rate (1 - test accuracy): 0.3448 -> 0.1255Hurt
use BAPerturb instead of FedPerturbmodel poisoning attack · CIFAR-10 (ResNet18), defense FedAvgattack success rate (1 - test accuracy): 0.7649 -> 0.9000Improved
use BAPerturb instead of FedPerturb across non-IID degreesmodel poisoning attack · CIFAR-10 (ResNet18), defense Multi-Krum, varying non-IIDattack success rate (1 - test accuracy)Improved
use BAPerturb instead of FedPerturb across local epochsmodel poisoning attack · Multi-Krum, varying malicious clients and epochsattack success rate (1 - test accuracy)Improved
apply BAPerturb against FLAME and RoseAggmodel poisoning attack · MNIST (LeNet)attack success rate (1 - test accuracy): 0.0165 and 0.0188No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy